Specific Correlation Between the Major Chromosome 10Q26 Haplotype Conferring Risk for Age-Related Macular Degeneration and the Expression of HTRA1

Specific correlation between the major chromosome 10q26 haplotype conferring risk for age-related macular degeneration and the expression of HTRA1

Sha-Mei Liao,1 Wei Zheng,1 Jiang Zhu,2 Casey A. Lewis,1 Omar Delgado,1 Maura A. Crowley,1 Natasha M. Buchanan,1 Bruce D. Jaffee,1 Thaddeus P. Dryja1

1Department of Ophthalmology; NIBR Informatics, Novartis Institutes for Biomedical Research, Cambridge, MA; 2Scientific Data Analysis, NIBR Informatics, Novartis Institutes for Biomedical Research, Cambridge, MA

Purpose: A region within chromosome 10q26 has a set of single nucleotide polymorphisms (SNPs) that define a haplotype that confers high risk for age-related macular degeneration (AMD). We used a bioinformatics approach to search for genes in this region that may be responsible for risk for AMD by assessing levels of gene expression in individuals carrying different haplotypes and by searching for open chromatin regions in the retinal pigment epithelium (RPE) that might include one or more of the SNPs. Methods: We surveyed the PubMed and the 1000 Genomes databases to find all common (minor allele frequency > 0.01) SNPs in 10q26 strongly associated with AMD. We used the HaploReg and LDlink databases to find sets of SNPs with alleles in linkage disequilibrium and used the Genotype-Tissue Expression (GTEx) database to search for correlations between genotypes at individual SNPs and the relative level of expression of the genes. We also accessed Encyclopedia of DNA Elements (ENCODE) to find segments of open chromatin in the region with the AMD-associated SNPs. Predicted transcription factor binding motifs were identified using HOMER, PROMO, and RegulomeDB software programs. Results: There are 34 polymorphisms within a 30-kb region that are in strong linkage disequilibrium (r2>0.8) with the reference SNP rs10490924 previously associated with risk for AMD. The expression of three genes in this region, PLEKHA1, ARMS2, and HTRA1 varies between people who have the low-AMD-risk haplotype compared with those with the high-AMD-risk haplotype. For PLEKHA1, 44 tissues have an expression pattern with the high-AMD-risk haplotype associated with low expression (rs10490924 effect size -0.43, p = 3.8 x 10-5 in ovary). With regard to ARMS2, the variation is most pronounced in testes: homozygotes with the high-AMD-risk haplotype express ARMS2 at lower levels than homozygotes with the low-AMD-risk haplotype; expression in heterozygotes falls in between (rs10490924 effect size -0.79, p = 7.5 x 10-24). For HTRA1, the expression pattern is the opposite; the high-AMD-risk haplotype has higher levels of expression in 27 tissues (rs10490924 effect size 0.40, p = 1.5 × 10−7 in testes). None of the other 22 genes within one megabase of rs10490924, or any gene in the entire genome, have mRNA expression levels that correlate with the high- AMD-risk haplotype. More than 100 other SNPs in the 10q26 region affect the expression of PLEKHA1 and ARMS2 but not that of HTRA1; none of these SNPs affects the risk for AMD according to published genome-wide association studies (GWASs). Two of the AMD-risk SNPs (rs36212732 and rs36212733) affect transcription factor binding sites in proximity to a DNase I hypersensitive region (i.e., a region of open chromatin) in RPE cells. Conclusions: SNPs in chromosome 10q26 that influence the expression of onlyPLEKHA1 or ARMS2 are not associated with risk for AMD, while most SNPs that influence the expression ofHTRA1 are associated with risk for AMD. Two of the AMD-risk SNPs affect transcription factor binding sites that may control expression of one of the linked genes in the RPE. These findings suggest that the variation in the risk for AMD associated with chromosome 10q26 is likely due to variation in HTRA1 expression. Modulating HTRA1 activity might be a potential therapy for AMD.

Age-related macular degeneration (AMD) is the leading some AMD-risk SNPs change the coding region of genes, cause of severe vision loss in older individuals. Single nucleo- most reside outside coding regions, suggesting potential tide polymorphisms (SNPs) in at least 34 loci influence the effects in regulating the expression of linked genes [2,3]. A risk for AMD [1], and the mechanisms by which these vari- set of closely linked SNPs on chromosome 10q26 is of special ants influence AMD are still being elucidated. Although interest since this chromosome has more influence on the risk for AMD than any other AMD region [1,4-7]; however, which gene in the region confers the risk remains unclear Correspondence to: Sha-Mei Liao, Department of Ophthalmology, [5,8-10]. Three genes are within 100 kilobase pairs (kb) of the Novartis Institutes for Biomedical Research, 22 Windsor Street, Cambridge, MA 02139, USA; Phone: (617) 871 4004; FAX: (617) SNPs associated with AMD. From the centromeric to telo- 871 5784; email: [email protected] meric ends of this region, the genes are pleckstrin homology Dr. Zheng now at GNS Healthcare, Cambridge, MA domain-containing family A member 1 (PLEKHA1; gene ID:

Table 1. The 34 polymorphisms in linkage disequilibrium that form an AMD-associated haplotype block.

Chr 10 position r2 LDlink/ D' LDlink/ SNP (GPCh37) REF/ALT Allele HaploReg HaploReg AMD References rs61871744 124,203,787 T/C 0.96/0.94 0.99/0.99 rs59616332 124,208,562 ATAAAC/- 0.96/NA 0.99/NA rs11200630 124,209,684 T/C 0.99/0.98 1/0.99 rs61871745 124,210,369 G/A 0.99/0.98 1/0.99 rs11200632 124,211,536 A/G 0.99/0.99 1/1 rs11200633 124,211,596 C/T 0.99/0.98 1/1 rs61871746 124,212,913 T/C 1/1 1/1 rs61871747 124,213,046 C/T 1/1 1/1 rs370974631 124,213,143 AA/- 0.93/NA 1/NA rs200227426 124,213,671 C/A 0.96/NA 0.99/NA rs201396317 124,213,674 C/A 0.96/NA 0.99/NA rs199637836 124,213,677 C/A 0.96/NA 0.99/NA rs11200634 124,213,680 C/A 0.96/NA 0.99/NA rs75431719 124,213,688 C/A 0.96/NA 0.99/NA 22, 23, 24, 25, 26, 27, 28, 36, rs10490924 124,214,448 G/T Reference SNP Reference SNP and additional 140 refs rs144224550 124,214,600 - /GT 1/NA 1/NA rs36212731 124,214,976 G/T 0.99/1 1/1 rs36212732 124,215,198 A/G 1/1 1/1 rs36212733 124,215,211 T/C 1/1 1/1 rs3750848 124,215,315 T/G 1/1 1/1 29 rs3750847 124,215,421 C/T 1/1 1/1 30 rs3750846 124,215,565 T/C 1/1 1/1 1, 29 rs566108895 124,216,823 G/T 0.91/NA 0.99/NA rs3793917 124,219,275 C/G 0.99/0.98 1/0.99 29, 31, 32, 33 rs3763764 124,220,061 A/G 0.98/0.98 0.99/0.99 17, 18, 19, 20, 21, 22, 23, and rs11200638 124,220,544 G/A 0.98/0.9 0.99/0.97 additional 100 refs rs1049331 124,221,270 C/T 0.97/0.96 0.99/0.99 34, 37 rs2293870 124,221,276 G/C,T NA/0.9 NA/0.95 34, 38, 39 rs2284665 124,226,630 G/T 0.96/0.94 0.99/0.98 35 rs60401382 124,227,624 C/T 0.84/0.82 0.98/0.97 rs11200643 124,229,203 C/T 0.82/0.8 0.96/0.94 rs58077526 124,230,024 A/C 0.91/0.89 0.96/0.95 rs932275 124,231,464 G/A 0.93/0.89 0.97/0.96 29 rs2142308 124,234,037 G/C 0.92/0.89 0.97/0.96

The SNP rs10490924 is the reference SNP used to obtain r2 and D’ values in both the HaploReg and LDlink websites. In the 1000 Genome project European population there are 34 polymorphisms, including rs10490924, in linkage disequilibrium using the inclusion criterion r2 ≥0.8 [11,12]. LDlink r2 refers to the strength of the association of the alleles at the queried SNP with rs10490924. LDlink D’ measures linkage disequilibrium normalized for allele frequency. NA = Not Available.

Figure 1. Diagram of the chromosome 10q26 region containing the PLEKHA1, ARMS2, and HTRA1 genes, as well as the locations of the SNPs that have been associated with risk for AMD by published GWASs and genetic studies.

59338, OMIM: 607772), age-related maculopathy-2 (ARMS2; 6. The mRNA levels are expressed as rank normalized gene gene ID: 387715, OMIM: 611313), and high temperature expression [15]. The effect size of an eQTL is the change in requirement protein A1 (HTRA1; gene ID: 5654, OMIM: the value of the standardized gene expression level with each 602194). It is even conceivable that a more distant gene actu- extra copy of the alternative (ALT) allele relative to the refer- ally confers the risk for AMD. ence allele, conditional to all other adjustments (gender, prob- We searched for associations between AMD-risk SNPs abilistic estimation of expression residuals factors, genotype in the 10q26 region and the expression of closely linked principal components, and genotyping platforms). Because genes. We also looked for open chromatin in the region in the gene expression levels of tissues have been transformed to RPE cells and transcription factor binding sites in the open a standard normal distribution (mean of 0 and standard devia- chromatin since such regions often correspond with regions tion of 1), the effect size is also equivalent to the change in that regulate transcription. the population standard deviation; that is, an effect size of 0.2 means a change in 0.2 standard deviations from the baseline level with both reference alleles. The effect size provides the METHODS variation in the strength of expression with positive numbers Linkage disequilibrium analysis: Using the online tools indicating higher mRNA levels in samples from people with HaploReg and LDlink, SNPs associated with AMD and other the minor allele compared to those with the major allele, and variants were placed in a haplotype block using Query SNP negative numbers indicating lower mRNA levels in samples rs10490924 and the linkage disequilibrium inclusion criterion with the minor allele. For the eQTL analysis, selected tissues 2 r ≥0.8 [11,12] in the 1000 Genome European population. were those for which high-quality mRNA results were avail- Transcriptome analysis: The GTEx database allows one able in at least 70 genotyped donors. to search for relationships between human genotypes and Search for open chromatin and transcription factor gene expression in specific tissues [13,14]. At the time of binding sites: We used the Encyclopedia of DNA Elements this analysis (January–October 2016), the GTEx project V6p (ENCODE) to search for DNase I sensitive sites; this database (GtexPortal) contained genotypes from 449 human adult has results from more than 125 cell types, including primary donors and whole genome transcription data from up to 53 cultures of RPE. We used three online programs that predict tissues from an overlapping set of 544 donors. Transcript transcription factor binding sites: HOMER, PROMO, and levels of PLEKHA1, ARMS2, and HTRA1 were included. The RegulomeDB. The consensus binding sequences are from steady-state mRNA level is measured as reads per kilobase JASPAR. of transcript per million mapped reads (RPKM). RPKM are Statistical analysis: The GTEx consortium database has normalized according to the number of sequencing reads and precalculated nominal eQTL p values for every human gene the read lengths. RPKM below 1 indicate levels of mRNA and all the informative SNPs analyzed. The p values for each expression that are difficult to distinguish from background SNP-gene pair were calculated using a two-tailed t test as noise. described at that site (GtexPortal). We considered associa- Association between genetic variants and mRNA expression: tions statistically significant when the p value was less than We searched for expression quantitative trait loci (eQTLs) 0.05. In some cases, as stated in the text, we made adjust- in the chromosome 10q26 region by looking for correlations ments for the multiple analyses. between SNP alleles and the expression of genes using the “Test your own eQTLs” option in the GTEx portal Version

Table 2. The rs10490924-T allele is associated with low levels (negative effect sizes) of PLEKHA1 gene expression in most tissues.

Expression (Mean Number Tissue P-value Effect Size RPKM) samples Brain - Nucleus accumbens (bg) 0.084 0.14 N/A 93 Stomach 0.45 0.038 20.678 170 Brain - Caudate (basal ganglia) 0.57 0.033 188.491 100 Brain - Hypothalamus 0.77 0.024 101.534 81 Heart - Left Ventricle 0.73 0.023 21.246 190 Brain - Frontal Cortex (BA9) 0.8 0.016 116.963 92 Brain - Putamen (basal ganglia) 0.82 0.016 193.492 82 Brain –Ant. cingulate cortex (BA24) 0.82 0.013 141.737 72 Heart - Atrial Appendage 0.93 0.0077 30.707 159 Small Intestine - Terminal Ileum 0.97 −0.0038 21.23 77 Testis 0.92 −0.0069 12.333 157 Liver 0.8 −0.024 52.72 97 Skin - Sun Exposed (Lower leg) 0.44 −0.032 62.651 302 Muscle - Skeletal 0.25 −0.041 6.046 361 Artery - Coronary 0.62 −0.043 145.955 118 Whole Blood 0.1 −0.046 1.153 338 Brain - Cerebellar Hemisphere 0.64 −0.052 28.95 89 Esophagus - Mucosa 0.35 −0.053 14.493 241 Pituitary 0.47 −0.067 105.377 87 Esophagus - Muscularis 0.14 −0.072 53.302 218 Thyroid 0.12 −0.074 42.25 278 Adipose - Visceral (Omentum) 0.067 −0.081 110.21 185 Skin -Not Sun Exposed (Suprapubic) 0.11 −0.084 60.731 196 Colon - Sigmoid 0.31 −0.086 74.374 124 Adipose - Subcutaneous 0.05 −0.088 126.917 298 Colon - Transverse 0.12 −0.089 36.749 169 Breast - Mammary Tissue 0.15 −0.089 114.907 183 Lung 0.0069 −0.11 36.291 278 Artery - Aorta 0.056 −0.11 268.567 197 Spleen 0.19 −0.14 26.428 89 Brain - Cortex 0.032 −0.15 118.249 96 Pancreas 0.051 −0.15 14.597 149 Artery - Tibial 0.0013 −0.18 139.083 285 Brain - Hippocampus 0.044 −0.18 129.512 81 Prostate 0.11 −0.18 37.874 87 Adrenal Gland 0.037 −0.21 36.181 126 Vagina 0.0068 −0.22 96.109 79 Uterus 0.083 −0.22 104.586 70 Nerve - Tibial 1.40E-08 −0.25 99.2 256 Brain - Cerebellum 0.003 −0.3 44.175 103

321 Molecular Vision 2017; 23:318-333 © 2017 Molecular Vision expression correlations) concentrated on the 25 SNPs in the GTEx database, comprising the ten that have been reported to be associated with risk for AMD and the 15 in strong linkage disequilibrium with those ten. We focused especially on the reference SNP rs10490924 because the strong linkage disequilibrium of the other SNPs with the reference SNP meant that results from any of them would approximately predict the results from the others. The present studies of the other SNPs in the haplotype confirmed that prediction. The three genes in 10q26 closest to the region of the AMD-associated SNPs are (centromeric to telomeric) PLEKHA1, ARMS2, and HTRA1. There is an intergenic region Figure 2. In the ovaries, women who are homozygous or hetero- of about 22 kilobases between PLEKHA1 and ARMS2 and zygous for the AMD-risk allele rs10490929-T have lower average an intergenic region of about 5 kilobases between ARMS2 PLEKHA1 mRNA levels compared with homozygotes with the reference allele G. Error bars indicate the standard error of the and HTRA1 (Figure 1). The 30-kb AMD-risk haplotype mean. The sample size (n) in each group is indicated above each stretches from the intergenic region between the PLEKHA1 genotype. and ARMS2 genes to the middle of the HTRA1 gene. The high-AMD-risk haplotype is associated with low expres- RESULTS sion of the PLEKHA1 gene: The PLEKHA1 gene is about 60 SNPs that define the AMD-risk haplotype in 10q26: There are kb in length and its termination codon, which is at the gene’s ten SNPs in a 17-kb region within 10q26 for which published telomeric boundary, is about 12 kb from the centromeric data indicate strong associations of alleles with risk for AMD boundary of the AMD-risk haplotype. PLEKHA1 is detect- (Table 1 and Figure 1) [1,16-39]. One of the SNPs, rs10490929, ably expressed in all 53 tissues in the GTEx database (Table 2 changes the ARMS2 coding sequence (G versus T, Ala69Ser). and Appendix 1) with an average expression level of about 9.5 The other nine SNPs are in strong linkage disequilibrium RPKM. In 32 tissues, the average PLEKHA1 expression was with rs10490929, and each has a correlation coefficient r2 lower in people with the high-AMD-risk haplotype defined ≥0.89 with rs10490924. Of these nine SNPs, rs3750846, by rs10490924-T, and the associated p values were below rs3750847, and rs3750848 are in the only ARMS2 intron, 0.05 in ten of those tissues (Table 2 and Appendix 2). Other rs3793917 and rs11200638 are located in the HTRA1 promoter SNPs in the AMD haplotype showed similar results, which region, rs1049331 and rs2293870 are synonymous variations was expected because of their strong linkage disequilibrium in the HTRA1 coding region, and rs2284665 and rs932275 are (Appendix 2). The effect was most striking in ovarian tissue in the first HTRA1 intron. We considered the set of alleles of (effect size −0.43, p = 3.8 × 10−5 for rs100490924; Figure 2). these ten SNPs, all from published studies showing associa- The effect appears to be semidominant with the expression tion with elevated risk for AMD, as a presumed haplotype level in heterozygotes falling between the TT and GG homo- that we refer to as the “high-AMD-risk” haplotype (Table zygotes. In the nine tissues that showed the opposite pattern 1). Although data from all ten SNPs are in the GTEx data- (higher expression levels in people with the high-AMD-risk base, we used rs10490924 as the reference SNP because it haplotype), the variation in mRNA levels among the geno- is the most commonly evaluated SNP in studies of AMD types was small and of no statistical significance, with all [16-28]. The HaploReg and LDlink websites indicate that the but one of the p values greater than 0.4. Thus, although the haplotype defined by the ten SNPs includes an additional 24 PLEKHA1 gene is more than 22 kb away from rs10490924 polymorphisms (21 SNPs and three insertion-deletion poly- and about 12 kb from SNP rs61871744 that defines the centro- morphisms) extending across a region of about 30 kb. Many meric boundary of the AMD-risk haplotype, the expression of of these 24 additional polymorphisms were not specifically this gene in many tissues is influenced by sequences within evaluated in most previous genetic evaluations of AMD, but the haplotype. they likely have some association with risk for AMD because The high-AMD-risk haplotype is associated with low expres- they have strong allelic correlations with rs10490924, with sion of ARMS2: The ARMS2 transcriptional unit is about 2 some having perfect correlations (i.e., r = 1; Table 1). Of 2 kb and is completely contained within the 30-kb region these 24 additional polymorphisms, 15 are included in the containing the AMD-risk SNPs. In the GTEx transcriptome GTEx database. The present analysis of eQTLs (SNP-gene data set, ARMS2 mRNA is not detected in 18 tissues and is at

322 Molecular Vision 2017; 23:318-333 © 2017 Molecular Vision a low level in most of the other tissues, averaging only about effect appears to be semidominant with the expression level 0.2 RPKM and never above 1.0 RPKM except testes that have in heterozygotes falling between the TT and GG homozy- a level of 3.0 RPKM (Table 3 and Appendix 3). In the testes, gotes. As expected because of strong linkage disequilibrium, homozygotes with the rs10490929-T allele (a marker of the another high-AMD-risk allele in the region (rs1120638-A), high-AMD-risk haplotype) generally have lower ARMS2 which is in the same haplotype as the rs10490924-T allele, is mRNA levels than homozygotes with the low-AMD-risk also associated with a low ARMS2 mRNA level in the testes G allele (effect size −0.79, p = 7.5 × 10−24; Figure 3). The (effect size −0.76, p = 8.3 × 10−20; Appendix 2). Eight other

Table 3. The rs10490924-T allele is associated with low levels (negative effect sizes) of ARMS2 gene expression in most tissues.

Expression (Mean Number Tissue P-Value Effect Size RPKM) Samples Brain - Anterior cingulate cortex (BA24) 0.95 0.012 N/A 72 Muscle - Skeletal 0.98 0.0017 0 361 Brain - Caudate (basal ganglia) 0.93 −0.015 0.057 100 Nerve - Tibial 0.71 −0.04 0.059 256 Liver 0.74 −0.048 0.099 97 Adipose - Subcutaneous 0.49 −0.056 0.113 298 Brain - Frontal Cortex (BA9) 0.76 −0.063 0.079 92 Breast - Mammary Tissue 0.51 −0.082 0.07 183 Colon - Transverse 0.18 −0.12 0.054 169 Stomach 0.1 −0.15 0.036 170 Thyroid 0.086 −0.16 0.051 278 Pituitary 0.33 −0.16 0.382 87 Brain - Putamen (basal ganglia) 0.37 −0.17 0.06 82 Heart - Left Ventricle 0.075 −0.18 0.055 190 Lung 0.098 −0.18 0.033 278 Prostate 0.35 −0.19 0.082 87 Adipose - Visceral (Omentum) 0.022 −0.2 0.116 185 Esophagus - Mucosa 0.069 −0.2 0.136 241 Brain - Cortex 0.24 −0.23 0.084 96 Brain - Hippocampus 0.23 −0.24 0.075 81 Vagina 0.088 −0.25 N/A 79 Heart - Atrial Appendage 0.042 −0.27 0.022 159 Artery - Tibial 0.00041 −0.35 0.099 285 Ovary 0.014 −0.35 0.472 85 Colon - Sigmoid 0.005 −0.4 0.152 124 Skin - Not Sun Exposed (Suprapubic) 0.0003 −0.41 0.115 196 Esophagus - Muscularis 0.000063 −0.45 0.136 218 Small Intestine - Terminal Ileum 0.0079 −0.45 0.044 77 Skin - Sun Exposed (Lower leg) 7.00E-07 −0.46 N/A 302 Artery - Aorta 0.00007 −0.47 0.109 197 Uterus 0.0049 −0.5 0.184 70 Brain - Nucleus accumbens (basal ganglia) 0.003 −0.54 0.05 93 Artery - Coronary 5.50E-06 −0.56 0.094 118 Brain - Hypothalamus 0.00019 −0.67 0.089 81 Testis 7.50E-24 −0.79 3.012 157

323 Molecular Vision 2017; 23:318-333 © 2017 Molecular Vision haplotype that we evaluated showed a similar pattern in HTRA1 mRNA levels in human tissues (Appendix 2). In one tissue (prostate) the opposite association was observed with an effect size of -0.23 (p = 0.02) with rs10490924. The high-AMD-risk haplotype has no detectable effect on the expression of distant genes: To test whether the high- AMD-risk haplotype could influence the expression of genes even more distant than PLEKHA1, ARMS2, or HTRA1, we examined the transcription of the 22 genes located within 1 megabase of the PLEKHA1-ARMS2-HTRA1 cluster in all 44 human tissues in the GTEx database. Of the 22 genes, 16 were detectably transcribed in at least one tissue. There was Figure 3. In the testes, men who are homozygous or heterozygous no statistically significant association between alleles at the for the AMD-risk allele rs10490929-T have lower average ARMS2 reference SNP rs10490924 and the expression of any of these mRNA levels compared with homozygotes with the reference allele G. Error bars indicate the standard error of the mean. The sample 16 genes in any of the 44 tissues. In fact, searching through size (n) in each group is indicated above each genotype. the entire transcriptomes of all 44 tissues in the GTEx database with the criterion of a false discovery rate (FDR) AMD-associated variants in this region that we evaluated are of less than or equal to 0.5, none of the AMD-associated associated with a similarly low expression of ARMS2 in the SNPs appeared to be eQTLs for any gene on any chromo- testes (Appendix 2). some except PLEKHA1, ARMS2, and HTRA1 (Table 5). Thus, the 10q26 high-AMD-risk haplotype appears to affect Among tissues other than the testes, 34 express detect- only PLEKHA1, ARMS2, and HTRA1, and it is unlikely that able but low levels of ARMS2 and thus provide poor statis- the high-AMD-risk haplotype has a distant eQTL or inter- tical power (Table 3). In 32 of the 34 tissues, the ARMS2 chromosome effect on gene expression. transcript levels were numerically lower in homozygotes with the rs10490929-T allele compared with heterozygotes Most SNPs from 10q26 that influence the expression of or G-allele homozygotes, a trend that is concordant with the PLEKHA1 and ARMS2 have not been associated with results from the testes (Table 3). The difference in expression AMD: Many more SNPs in 10q26 influence the expression between people with the high-AMD-risk T allele versus the of PLEKHA1 and ARMS2 than HTRA1 (Figure 5A). For low-AMD-risk G allele reached p values less than or equal example, the expression of PLEKHA1 across 12 tissues is to 0.05 in 14 of the tissues, such as sun-exposed skin (p = 7.0 significantly influenced by more than 254 SNPs, with “signif- × 10 −7) and coronary artery (p = 5.5 × 10−6). Analysis of the icance” defined by the GTEx database significance criterion other SNPs in the AMD-risk haplotype that we evaluated of an FDR of 5% or less. Of the 254 SNPs, 28 also affect the showed a similar pattern (Appendix 2). expression of HTRA1, while the remaining 226 SNPs affect the expression of PLEKHA1 but not HTRA1. None of the 226 The high-AMD-risk haplotype is associated with high expres- has been found to influence the risk for AMD. Only 25 of the sion of the HTRA1 gene: The SNP rs2142308, which forms 254 SNPs that influence the expression of PLEKHA1 also the telomeric boundary of the AMD-risk haplotype, is inside modulate the risk for AMD, and all 25 are among those that intron 1 of HTRA1. HTRA1 mRNA is ubiquitously expressed modulate HTRA1 expression. The expression of ARMS2 in with RPKM values ranging from 1 (whole blood) to 269 (the 13 tissues is influenced by 192 SNPs extending across 855 aorta) and averages about 400-fold higher than ARMS2 and kb (Figure 5A and Appendix 2). Of the 192 SNPs, 35 also 8.5-fold higher than PLEKHA1 (81.2 RPKM versus 0.2 or 9.5 affect the expression of HTRA1, while the remaining 157 RPKM, respectively; Table 4 and Appendix 4). In 27 tissues, affect the expression of ARMS2 but not HTRA1. None of the the average HTRA1 expression was higher in people with the 157 has been found to influence the risk for AMD. Only 25 high-AMD-risk haplotype defined by rs10490924-T, and the of the 192 SNPs that influence the expression of ARMS2 also effect was associated with p values of less than 0.05 in six modulate the risk for AMD, and all 25 are among those that of those tissues (Table 4 and Appendix 2). The effect was modulate HTRA1 expression. Thus, more than 85% of the most striking in the testes (effect size 0.4, p = 1.5 × 10−7 for SNPs that influence the expression of ARMS2 and PLEKHA1 rs10490924; Figure 4 and Table 4). In one tissue (prostate), are not associated with risk for AMD. Ninety-nine of these the opposite association was observed with an effect size of SNPs have high minor allele frequencies (higher than 0.2), −0.23 (p = 0.02). Analysis of the other SNPs in the AMD-risk

Table 4. The rs10490924-T allele is associated with low levels (positive effect sizes) of HTRA1 gene expression in most tissues.

Expression (Mean Number Tissue P-Value Effect Size RPKM) Samples Testis 1.50E-07 0.4 12.333 157 Pituitary 0.016 0.28 105.377 87 Artery - Coronary 0.043 0.16 144.948 118 Artery - Aorta 0.017 0.13 268.567 197 Liver 0.12 0.12 52.72 97 Breast - Mammary Tissue 0.074 0.11 114.907 183 Whole Blood 0.029 0.1 1.153 338 Lung 0.032 0.098 36.291 278 Adipose - Subcutaneous 0.056 0.094 126.917 298 Small Intestine - Terminal Ileum 0.24 0.093 21.23 77 Vagina 0.12 0.09 96.109 79 Adrenal Gland 0.36 0.082 36.291 126 Nerve - Tibial 0.26 0.077 99.2 256 Colon - Transverse 0.074 0.066 36.749 169 Spleen 0.45 0.06 26.428 89 Brain - Hippocampus 0.63 0.051 129.512 81 Heart - Atrial Appendage 0.44 0.043 30.707 159 Thyroid 0.32 0.039 42.25 278 Esophagus - Mucosa 0.46 0.032 14.493 241 Muscle - Skeletal 0.34 0.026 6.046 361 Skin - Sun Exposed (Lower leg) 0.64 0.019 62.815 302 Artery - Tibial 0.78 0.01 139.083 285 Adipose - Visceral (Omentum) 0.86 0.0092 110.21 185 Heart - Left Ventricle 0.88 0.0084 21.246 190 Brain - Nucleus accumbens (bg) 0.97 0.0036 132.309 93 Pancreas 0.98 0.0024 14.597 149 Ovary 0.99 0.0012 211.834 85 Stomach 0.78 −0.012 20.678 170 Brain - Hypothalamus 0.83 −0.026 101.534 81 Brain - Cortex 0.68 −0.03 118.249 96 Brain - Caudate (basal ganglia) 0.68 −0.037 188.491 100 Brain - Anterior cingulate cortex (BA24) 0.29 −0.069 141.737 72 Esophagus - Muscularis 0.13 −0.081 53.302 218 Colon - Sigmoid 0.29 −0.089 74.374 124 Brain - Frontal Cortex (BA9) 0.21 −0.092 116.804 92 Brain - Cerebellum 0.18 −0.099 44.175 103 Skin - Not Sun Exposed (Suprapubic) 0.12 −0.1 60.731 196 Brain - Putamen (basal ganglia) 0.18 −0.1 193.492 82 Brain - Cerebellar Hemisphere 0.15 −0.12 28.95 89 Uterus 0.073 −0.21 104.586 70 Prostate 0.017 −0.23 37.874 87

325 Molecular Vision 2017; 23:318-333 © 2017 Molecular Vision 30-kb AMD-risk region, there are three DNase I hypersensitive sites in RPE cells (Table 6). We evaluated whether any of the 34 polymorphisms in the AMD-risk haplotype, including the 25 in the GTEx database and the nine additional polymorphisms in the LDlink database, were within the DNase I hypersensitive sites. The site with the most DNase I hyper- sensitivity is a 170-bp segment extending from chromosome 10 positions 124,215,021-124,215,190 (Figure 5B). This site is 5.8 kb upstream of the transcription start site of HTRA1 and within the intron of ARMS2. Using the transcription-factor- binding-motif-prediction software HOMER, PROMO, and RegulomeDB, we searched for consensus binding sequences Figure 4. In the testes, men who are homozygous or heterozygous for transcription factors in the open chromatin region that for the AMD-risk allele rs10490929-T have higher average higher might include AMD-risk SNPs. The results implicated two HTRA1 mRNA levels compared with homozygotes with the refer- AMD-risk SNPs, rs36212732 and rs36212733, which are 8 bp ence allele G. Error bars indicate the standard error of the mean. and 21 bp, respectively, away from the telomeric boundary The sample size (n) in each group is indicated above each genotype. of the hypersensitive site (Figure 6). In comparison with the low-AMD-risk haplotype, the high-AMD-risk haplotype and 35 have a high effect size (absolute value higher than 1; loses sites for transcription factors YY1, LHX2, LHX3, Appendix 2). Some of these SNPs were definitely included NKX6–1, ALX1, and ALX3, and it creates a site for the in published genome-wide association studies (GWASs). For transcription factor c-MYB (Figure 6). These transcription example, 20 SNPs that affect PLEKHA1 and ARMS2 but factors are expressed in human RPE cells [41,42]. not HTRA1 are in the Illumina Human 610-Quad BeadChip used by Fritsch et al. who did not report an association with The other two open chromatin sites in this region are AMD with any of these SNPs [40]. It is likely that among the less than one tenth as sensitive to DNase I (Table 6). One numerous, large GWASs conducted on patients with AMD, extends from 124,220,906-124,222,950 and contains two some of these SNPs would have been identified as influencing AMD-risk SNPs (rs1049331 and rs2293870), but these poly- risk for AMD if they actually had an effect on risk for AMD. morphisms do not affect any potential transcription factor binding sites according to the HOMER, PROMO, and Regu- For HTRA1, only 41 SNPs influence its expression. lomeDB programs. The other open chromatin site extends Thirty-four of these 41 SNPs are concentrated in the 30-kb from 124,228,506-124,228,935 and has no AMD-associated AMD haplotype region, including all 25 SNPs in the AMD- SNPs within or nearby. risk haplotype (Figure 5B). All AMD-risk SNPs that influence the expression of PLEKHA1 or ARMS2 also influence DISCUSSION the expression of HTRA1. We used the GTEx genotype-tissue expression database to Two AMD-risk SNPs alter predicted transcription factor explore the effects of SNPs that influence the susceptibility binding sites: The ENCODE database of open chromatin for AMD on the expression of nearby genes in the 10q26 regions includes data from the RPE, a monolayer of cells ARMS2-HTRA1 region. The AMD-risk alleles at ten SNPs that plays a key role in the pathogenesis of AMD. Within the

Table 5. Single tissue significant cis-eQTLs for AMD-associated SNPs rs10490924.

Gencode ID Gene Symbol P-Value Effect Size Tissue ENSG00000254636.1 ARMS2 7.50E-24 −0.79 Testis ENSG00000107679.10 PLEKHA1 1.40E-08 −0.25 Nerve - Tibial ENSG00000166033.7 HTRA1 1.50E-07 0.4 Testis ENSG00000254636.1 ARMS2 7.00E-07 −0.46 Skin - Sun Exposed (Lower leg)

The table provides the results from a search of GTEx for the entire list of genes that are influenced by the reference SNPrs10490924 using the GTEx default criteria for significance. Note that only 3 genes appear, indicating that no other gene anywhere in chromosome 10q26 nor any gene in the GTEx database throughout the human genome is influenced by the reference SNP. Similar results appear for all of the 25 GTEx SNPs in the AMD-risk haplotype (data not shown), as expected since they are all in strong linkage disequilibrium.

326 Molecular Vision 2017; 23:318-333 © 2017 Molecular Vision in this region are in strong linkage disequilibrium, and there in gene expression (rather than a change in the transcribed are 24 additional nearby SNPs or insertion-deletion polymor- protein), then the risk is due to variation in the expression of phisms that are similarly correlated, 15 of which are in the one of these three genes. GTEx database. Based on results from 25 of these SNPs in The GTEx data additionally provide evidence that varia- the GTEx database, high-AMD-risk alleles are associated tions in ARMS2 and PLEKHA1 expression are less likely than with lower levels of ARMS2 and PLEKHA1 mRNA and HTRA1 to influence risk for AMD. Hundreds of additional higher levels of HTRA1 mRNA than low-AMD-risk alleles SNPs in this region influence the expression of ARMS2 and in many human tissues. The AMD-risk SNPS do not influ- PLEKHA1 but not HTRA1. None of those additional SNPs ence the level of expression of any other gene in this region or has been associated with the risk for AMD in any published anywhere in the human genome with statistical significance human genetics study although some of these SNPs were after adjustment for the multiple comparisons. Thus, if the included in those studies. However, most of the SNPs (25/41) risk for AMD conferred by this region is due to variation

Figure 5. The locations of single nucleotide polymorphisms (SNPs) that influence the expression of PLEKHA1, ARMS2, or HTRA1 (i.e., cis-expression quantitative trait loci (eQTLs)). A: The top of the figure shows the intron–exon structures of the PLEKHA1, ARMS2, and HTRA1 genes. Under the schematic genome segment are three graphs with each graph showing the locations of SNPs that affect the mRNA expression (i.e., eQTLs) for each respective gene. The x-axis has the base pair locations, with the scale numbers at the bottom of the three graphs based on human genome reference GRCh37/hg19. The dots are colored so they correspond to panel B (green = PLEKHA1, blue = ARMS2, red = HTRA1). To be an eQTL on this graph, an SNP must meet the threshold for significance as calculated by the GTEx software (a false discovery rate of less than 0.5) in at least one tissue. The effect size (y-axis) is the maximum effect size across all tissues, with positive values meaning that the minor (alternative) allele is associated with higher expression. There are some eQTLs (75 for PLEKHA1, 44 for ARMS2, and four for HTRA1) off the ends of these graphs that are not included because of size limitations. B: Expanded view of the eQTLs in the region of the AMD-risk haplotype. The x-axis in this view is not linear. Black arrows point to the ten SNPs that have been reported in genome-wide association studies (GWASs) and genetic studies as associated with risk for AMD. Unfilled arrows are SNPs in strong linkage disequilibrium with the reported AMD-risk SNPs (linkage disequilibrium r2 >0.8) and that therefore are highly likely to be associated with risk for AMD but have not been reported as such in GWASs, perhaps because they have not been included in those studies. Note that all AMD-risk SNPs are associated with high HTRA1, low PLEKHA1, and low ARMS2 mRNA expression. Note also that there are 18 SNPs within the region of the AMD-risk haplotype in B and more than 100 that are away from it (shown in A) that affect the expression of PLEKHA1 and ARMS2 (denoted by SNPs with green or blue dots) but that have never been reported to influence risk for AMD. The location of the open chromatin region in the RPE with the highest DNase I sensitivity is indicated as a horizontal bar. Not shown is the ins/ del polymorphism esv2663177 (del443/ins54) which is between rs3750846 and rs3793917.

Figure 6. Changes in predicted transcription factor binding sites by SNPs associated with risk for AMD. The single nucleotide polymorphisms (SNPs) rs36212732 and rs36212733 are 8–21 bp away from the telomeric boundary of an open chromatin region with the high DNase I sensitivity in RPE cells. The low-AMD-risk haplotype has the base “A” at rs36212732 and the base “T” at rs36212733, while the high-AMD-risk allele has the bases “G” and “C” at the respective locations. In comparison with the low-AMD-risk haplotype, the high-AMD-risk haplotype has a predicted reduction in the binding of the transcription factor YY-1, LHX2, LHX3, NKX6–1, ALX1, and ALX3, and a predicted increase in the binding of the factor c-MYB, as predicted by the software HOMER, PROMO, and RegulomeDB. The consensus YY-1, LHX2, LHX3, NKX6–1, ALX1, ALX3, and c-MYB binding sequences are from the JASPAR website. that influence HTRA1 expression are associated with risk for in the primary structure of an encoded protein. Only one AMD either from direct evidence from GWASs or because such polymorphism is known among the three candidates. It the SNPs are highly correlated with directly implicated involves the reference SNP rs10490924, which is a missense SNPs. Two other items provide further support for disre- polymorphism (Ala69Ser) that affects the ARMS2 coding garding ARMS2 as an AMD-susceptibility gene: 1) ARMS2 sequence. The Ala69 ARMS2 allele corresponds to low risk mRNA and protein are expressed at extremely low levels in for AMD and high ARMS2 expression while the Ser69 allele eye tissues [9,18]; and 2) human genetics studies of AMD corresponds to high risk for AMD and low ARMS2 expres- indicate that a specific SNP that creates a nonsense mutation sion. Evidence against this polymorphism as the basis for (Arg38End) in ARMS2 is associated with low risk for AMD risk for AMD is as follows. If expression of Ala69-ARMS2 [5,10,43-45]. In short, HTRA1 is the most likely candidate protects against AMD, one would expect that loss of ARMS2 gene for risk for AMD in this region, and if so, elevated expression would always be associated with elevated risk for expression of HTRA1 likely increases risk for AMD. AMD. However, a separate ARMS2 variant, the nonsense The present analysis was based heavily on the idea that change Arg38End, would be expected to produce no func- the level of expression of a gene on 10q26 determines risk tional ARMS2, yet the variant has been found to confer low for AMD. An alternative explanation is that there is a change risk for AMD. The possibility remains that the Ser69 allele

Table 6. DNase I hypersensitive clusters at chromosome 10q26.13 in HRPEpiC* and changes in transcription factor binding sites from low-AMD-risk reference alleles to high-AMD-risk alternate alleles.

Transcription factor SNPs in the Transcription factor motif in motif in alternate Position Genomic size Signal AMD haplotype reference allele allele rs36212731 chr10:124215021– 170 163 rs36212732 YY1, LHX2, LHX3, NKX6–1, 124215190 rs36212733 ALX1, ALX3 c-MYB chr10:124220906– rs1049331 2045 12 124222950 rs2293870 chr10:124228506– 430 13 124228935

* HRPEpiC: Human Retinal Pigment Epithelial Cells

328 Molecular Vision 2017; 23:318-333 © 2017 Molecular Vision promotes the development of AMD because of some toxic is conceivable that variation in these two SNPs is the funda- effect of the Ser69-ARMS2 protein. This mechanism remains mental cause for risk for AMD in 10q26 and that the changes a possibility, but we feel it is unlikely because the Ser69 in transcription factor affinity mediated by the SNP alleles variant is expressed at low levels across all evaluated tissues. cause the variation in the expression of the HTRA1 gene In particular, a recent report showed that the Ser69-ARMS2 located 5.8 kb away. protein could not be detected in monocytes from patients There are other possible mechanisms for the variation in carrying the homozygous risk for AMD rs10490924-T variant the expression HTRA1 or other genes due to the AMD-risk [46]. haplotype, three of which are the following. 1) AMD-risk Support for HTRA1 as the risk for AMD factor comes SNPs, such as rs10490924, are in strong genetic linkage from reports of two- to threefold higher HTRA1 expression disequilibrium with the insertion/deletion polymorphism in eyes with AMD [10,17,47-51], although other groups report del443ins54 in the 3′ untranslated region of ARMS mRNA no effect [28,45,52-55]. Some support for low HTRA1 expres- [29]. It is possible that the del443ins54 polymorphism intro- sion protecting against AMD comes from patients who lack duces a conformational change in chromatin thus affecting the HTRA1 due to recessive, null mutations. Such patients have expression of genes in 10q26. However, the recently reported cerebral arteries with small lumens and thick walls with minimal region responsible for risk for AMD does not include reduplicated elastic laminas. No AMD has been observed in this polymorphism, making it unlikely that it modulates risk such patients [56]. for AMD [58]. 2) The reported pattern of DNA methylation There are weaknesses in the present analysis. Although in the promoter region of ARMS2 correlates with the risk the GTEx database includes six tissues with a strong correla- for AMD allele rs10490924-T [59]. However, the variation tion between high-AMD-risk SNP alleles and higher HTRA1 in methylation would more likely affect the expression of expression level and 21 others have a trend in the same ARMS2 rather than HTRA1 or PLEKHA1. 3) It is possible direction, one tissue (prostate) showed a correlation in the that transcriptional activity of the ARMS2 or PLEKHA1 opposite direction. It would be ideal to have ocular tissues genes might interfere with transcription of the nearby for evaluation, but unfortunately, no data from ocular tissues HTRA1. Chimeric transcripts starting from PLEKHA1 and are available in the GTEx database. It is still unclear to what ending in ARMS2 were recently reported [5]. The reduction in extent systemic factors influence one’s risk for AMD [57]. ARMS2 and PLEKHA1 transcription by the high-AMD-risk It is conceivable that variation in the systemic expression variants might consequently allow more HTRA1 mRNA to of HTRA1, not its local ocular expression, is responsible for be transcribed. This indirect effect on HTRA1 gene expres- increasing the risk for AMD. sion may explain why fewer tissues with increased HTRA1 gene expression reached statistical significance compared to A potential mechanism for the variation in the expres- ARMS2 in the GTEx database analysis. sion HTRA1 due to the AMD-risk haplotype involves an open chromatin (DNase I-sensitive) region in the RPE that APPENDIX 1. PLEKHA1 MRNA LEVEL IS we found in the ENCODE database. This region is 8–21 bp EXPRESSED IN ALL TISSUES EVALUATED IN THE away from the AMD-risk SNPs rs36212732 and rs36212733. GTEX PROJECT. Specifically, the change from the low-AMD-risk allele to the high-AMD-risk allele switches transcription factor binding To access the data, click or select the words “Appendix 1.” from YY-1, LHX2, LHX3, ALX1, ALX3, or NKX6–1 to c-MYB (Figure 6). APPENDIX 2. MAXIMUM EFFECT SIZES FOR PLEKHA1, ARMS2, AND HTRA1 CIS-EQTLS. A recently published evaluation of the ARMS2-HTRA1 region provides additional evidence for the importance of the To access the data, click or select the words “Appendix 2.” In open chromatin region and the SNPs near it that potentially the column corresponding to each gene, the number in a box affect transcription factor binding sites [58]. Based on 33,000 is the maximum eQTL effect seen across all tissues in the AMD cases and controls, the interval most likely responsible GTEx database. We excluded eQTLs that are not statistically for risk for AMD was narrowed down to a 7136-bp segment significant (false discovery rate > .05). In the rare instances bounded by SNPs rs11200630 and esv2663177. This interval where one tissue has an effect in the opposite direction from is within the AMD haplotype we defined and contains 13 the majority of tissues, only the maximum effect from the of the 25 SNPs. This interval includes the open chromatin majority of tissues is included. region we uncovered, as well as the SNPs rs36212732 and rs36212733 that affect transcription factor binding sites. It

329 Molecular Vision 2017; 23:318-333 © 2017 Molecular Vision APPENDIX 3. ARMS2 MRNA LEVEL IS HIGHEST IN REFERENCES TESTIS COMPARED TO OTHER TISSUES IN THE 1. Fritsche LG, Igl W, Bailey JN, Grassmann F, Sengupta S, GTEX PROJECT. Bragg-Gresham JL, Burdon KP, Hebbring SJ, Wen C, Gorski M, Kim IK, Cho D, Zack D, Souied E, Scholl HP, Bala E, To access the data, click or select the words “Appendix 3.” Lee KE, Hunter DJ, Sardell RJ, Mitchell P, Merriam JE, Many tissues have expression levels below 1 RPKM which Cipriani V, Hoffman JD, Schick T, Lechanteur YT, Guymer are difficult to distinguish from background noise and may RH, Johnson MP, Jiang Y, Stanton CM, Buitendijk GH, Zhan indicate no substantive expression X, Kwong AM, Boleda A, Brooks M, Gieser L, Ratnapriya R, Branham KE, Foerster JR, Heckenlively JR, Othman MI, Vote BJ, Liang HH, Souzeau E, McAllister IL, Isaacs T, Hall APPENDIX 4. HTRA1 MRNA LEVEL IS EXPRESSED J, Lake S, Mackey DA, Constable IJ, Craig JE, Kitchner TE, IN ALL TISSUES EVALUATED IN THE GTEX Yang Z, Su Z, Luo H, Chen D, Ouyang H, Flagg K, Lin D, PROJECT. Mao G, Ferreyra H, Stark K, von Strachwitz CN, Wolf A, To access the data, click or select the words “Appendix 4.” Brandl C, Rudolph G, Olden M, Morrison MA, Morgan DJ, Schu M, Ahn J, Silvestri G, Tsironi EE, Park KH, Farrer LA, Orlin A, Brucker A, Li M, Curcio CA, Mohand-Said ACKNOWLEDGMENTS S, Sahel JA, Audo I, Benchaboune M, Cree AJ, Rennie CA, The authors thank Eric Marshall for assisting with the GTEx Goverdhan SV, Grunin M, Hagbi-Levi S, Campochiaro P, Katsanis N, Holz FG, Blond F, Blanche H, Deleuze JF, Igo RP data analysis. The GTEx website asks for the following Jr, Truitt B, Peachey NS, Meuer SM, Myers CE, Moore EL, acknowledgment: “The Genotype-Tissue Expression (GTEx) Klein R, Hauser MA, Postel EA, Courtenay MD, Schwartz Project was supported by the Common Fund of the Office of SG, Kovach JL, Scott WK, Liew G, Tan AG, Gopinath B, the Director of the National Institutes of Health. Additional Merriam JC, Smith RT, Khan JC, Shahid H, Moore AT, funds were provided by the NCI, NHGRI, NHLBI, NIDA, McGrath JA, Laux R, Brantley MA Jr, Agarwal A, Ersoy L, Caramoy A, Langmann T, Saksens NT, de Jong EK, NIMH, and NINDS. Donors were enrolled at Biospecimen Hoyng CB, Cain MS, Richardson AJ, Martin TM, Blangero Source Sites funded by NCI\SAIC-Frederick, Inc. (SAIC-F) J, Weeks DE, Dhillon B, van Duijn CM, Doheny KF, Romm subcontracts to the National Disease Research Interchange J, Klaver CC, Hayward C, Gorin MB, Klein ML, Baird PN, (10XS170), Roswell Park Cancer Institute (10XS171), and den Hollander AI, Fauser S, Yates JR, Allikmets R, Wang Science Care, Inc. (X10S172). The Laboratory, Data Analysis, JJ, Schaumberg DA, Klein BE, Hagstrom SA, Chowers I, and Coordinating Center (LDACC) was funded through a Lotery AJ, Leveillard T, Zhang K, Brilliant MH, Hewitt AW, Swaroop A, Chew EY, Pericak-Vance MA, DeAngelis M, contract (HHSN268201000029C) to The Broad Institute, Inc. Stambolian D, Haines JL, Iyengar SK, Weber BH, Abecasis Biorepository operations were funded through an SAIC-F GR, Heid IM. A large genome-wide association study of age- subcontract to Van Andel Institute (10ST1035). Additional related macular degeneration highlights contributions of rare data repository and project management were provided by and common variants. Nat Genet 2016; 48:134-43. [PMID: SAIC-F (HHSN261200800001E). The Brain Bank was 26691988]. supported by supplements to University of Miami grants 2. Consortium EP. A user’s guide to the encyclopedia of DNA DA006227 & DA033684 and to contract N01MH000028. elements (ENCODE). PLoS Biol 2011; 9:e1001046-[PMID: 21526222]. Statistical Methods development grants were made to the University of Geneva (MH090941 & MH101814), the 3. Mu XJ, Lu ZJ, Kong Y, Lam HY, Gerstein MB. Analysis of genomic variation in non-coding elements using population- University of Chicago (MH090951, MH090937, MH101820, scale sequencing data from the 1000 Genomes Project. MH101825), the University of North Carolina - Chapel Hill Nucleic Acids Res 2011; 39:7058-76. [PMID: 21596777]. (MH090936 & MH101819), Harvard University (MH090948), 4. Hu Z, Xie P, Ding Y, Yuan D, Liu Q. Association between Stanford University (MH101782), Washington University variants A69S in ARMS2 gene and response to treatment of St Louis (MH101810), and the University of Pennsylvania exudative AMD: a meta-analysis. Br J Ophthalmol 2015; (MH101822). The data used for the analyses described in 99:593-8. [PMID: 25185256]. this manuscript were obtained from the GTEx Portal from 5. Kortvely E, Ueffing M. Gene Structure of the 10q26 Locus: 11/01/2015 to 12/04/2016. Parts of the data (Figure 2-4 and A Clue to Cracking the ARMS2/HTRA1 Riddle? Adv Exp Table 2-4) have been presented at ARVO 2016. All work was Med Biol 2016; 854:23-9. [PMID: 26427389]. performed at Novartis. 6. Jakobsdottir J, Conley YP, Weeks DE, Mah TS, Ferrell RE, Gorin MB. Susceptibility genes for age-related maculopathy on chromosome 10q26. Am J Hum Genet 2005; 77:389-407. [PMID: 16080115].

330 Molecular Vision 2017; 23:318-333 © 2017 Molecular Vision 7. Fisher SA, Abecasis GR, Yashar BM, Zareparsi S, Swaroop A, 19. Cameron DJ, Yang Z, Gibbs D, Chen H, Kaminoh Y, Jorgensen Iyengar SK, Klein BE, Klein R, Lee KE, Majewski J, Schultz A, Zeng J, Luo L, Brinton E, Brinton G, Brand JM, Bernstein DW, Klein ML, Seddon JM, Santangelo SL, Weeks DE, PS, Zabriskie NA, Tang S, Constantine R, Tong Z, Zhang K. Conley YP, Mah TS, Schmidt S, Haines JL, Pericak-Vance HTRA1 variant confers similar risks to geographic atrophy MA, Gorin MB, Schulz HL, Pardi F, Lewis CM, Weber and neovascular age-related macular degeneration. Cell BH. Meta-analysis of genome scans of age-related macular Cycle 2007; 6:1122-5. [PMID: 17426452]. degeneration. Hum Mol Genet 2005; 14:2257-64. [PMID: 20. Yoshida T, DeWan A, Zhang H, Sakamoto R, Okamoto H, 15987700]. Minami M, Obazawa M, Mizota A, Tanaka M, Saito Y, 8. Wang G. Chromosome 10q26 locus and age-related macular Takagi I, Hoh J, Iwata T. HTRA1 promoter polymorphism degeneration: a progress update. Exp Eye Res 2014; 119:1-7. predisposes Japanese to age-related macular degeneration. [PMID: 24291204]. Mol Vis 2007; 13:545-8. [PMID: 17438519]. 9. Fritsche LG, Loenhardt T, Janssen A, Fisher SA, Rivera A, 21. Mori K, Horie-Inoue K, Kohda M, Kawasaki I, Gehlbach PL, Keilhauer CN, Weber BH. Age-related macular degeneration Awata T, Yoneya S, Okazaki Y, Inoue S. Association of the is associated with an unstable ARMS2 (LOC387715) mRNA. HTRA1 gene variant with age-related macular degeneration Nat Genet 2008; 40:892-6. [PMID: 18511946]. in the Japanese population. J Hum Genet 2007; 52:636-41. 10. Yang Z, Tong Z, Chen Y, Zeng J, Lu F, Sun X, Zhao C, Wang [PMID: 17568988]. K, Davey L, Chen H, London N, Muramatsu D, Salasar F, 22. Lu F, Hu J, Zhao P, Lin Y, Yang Y, Liu X, Fan Y, Chen B, Liao Carmona R, Kasuga D, Wang X, Bedell M, Dixie M, Zhao S, Du Q, Lei C, Cameron DJ, Zhang K, Yang Z. HTRA1 P, Yang R, Gibbs D, Liu X, Li Y, Li C, Li Y, Campochiaro variant increases risk to neovascular age-related macular B, Constantine R, Zack DJ, Campochiaro P, Fu Y, Li DY, degeneration in Chinese population. Vision Res 2007; Katsanis N, Zhang K. Genetic and functional dissection of 47:3120-3. [PMID: 17904186]. HTRA1 and LOC387715 in age-related macular degenera- 23. Leveziel N, Souied EH, Richard F, Barbu V, Zourdani A, tion. PLoS Genet 2010; 6:e1000836-[PMID: 20140183]. Morineau G, Zerbib J, Coscas G, Soubrane G, Benlian P. 11. Ward LD, Kellis M. HaploReg: a resource for exploring chro- PLEKHA1–LOC387715-HTRA1 polymorphisms and exuda- matin states, conservation, and regulatory motif alterations tive age-related macular degeneration in the French popula- within sets of genetically linked variants. Nucleic Acids Res tion. Mol Vis 2007; 13:2153-9. [PMID: 18079691]. 2012; 40:D930-4. [PMID: 22064851]. 24. Shuler RK Jr, Hauser MA, Caldwell J, Gallins P, Schmidt 12. Machiela MJ, Chanock SJ. LDlink: a web-based application S, Scott WK, Agarwal A, Haines JL, Pericak-Vance MA, for exploring population-specific haplotype structure and Postel EA. Neovascular age-related macular degeneration linking correlated alleles of possible functional variants. and its association with LOC387715 and complement factor Bioinformatics 2015; 31:3555-7. [PMID: 26139635]. H polymorphism. Arch Ophthalmol 2007; 125:63-7. [PMID: 13. Consortium GT. The Genotype-Tissue Expression (GTEx) 17210853]. project. Nat Genet 2013; 45:580-5. [PMID: 23715323]. 25. Ross RJ, Bojanowski CM, Wang JJ, Chew EY, Rochtchina E, 14. Consortium GT. Human genomics. The Genotype-Tissue Ferris FL 3rd, Mitchell P, Chan CC, Tuo J. The LOC387715 Expression (GTEx) pilot analysis: multitissue gene regulation polymorphism and age-related macular degeneration: repli- in humans. Science 2015; 348:648-60. [PMID: 25954001]. cation in three case-control samples. Invest Ophthalmol Vis Sci 2007; 48:1128-32. [PMID: 17325155]. 15. Bolstad BM, Irizarry RA, Astrand M, Speed TP. A comparison of normalization methods for high density oligonucleotide 26. Francis PJ, George S, Schultz DW, Rosner B, Hamon S, Ott J, array data based on variance and bias. Bioinformatics 2003; Weleber RG, Klein ML, Seddon JM. The LOC387715 gene, 19:185-93. [PMID: 12538238]. smoking, body mass index, environmental associations with advanced age-related macular degeneration. Hum Hered 16. Fritsche LG, Fariss RN, Stambolian D, Abecasis GR, Curcio 2007; 63:212-8. [PMID: 17347568]. CA, Swaroop A. Age-related macular degeneration: genetics and biology coming together. Annu Rev Genomics Hum 27. Kondo N, Honda S, Ishibashi K, Tsukahara Y, Negi A. Genet 2014; 15:151-71. [PMID: 24773320]. LOC387715/HTRA1 variants in polypoidal choroidal vasculopathy and age-related macular degeneration in a Japanese 17. Yang Z, Camp NJ, Sun H, Tong Z, Gibbs D, Cameron DJ, Chen population. Am J Ophthalmol 2007; 144:608-12. [PMID: H, Zhao Y, Pearson E, Li X, Chien J, Dewan A, Harmon J, 17692272]. Bernstein PS, Shridhar V, Zabriskie NA, Hoh J, Howes K, Zhang K. A variant of the HTRA1 gene increases suscepti- 28. Kanda A, Chen W, Othman M, Branham KE, Brooks M, bility to age-related macular degeneration. Science 2006; Khanna R, He S, Lyons R, Abecasis GR, Swaroop A. A 314:992-3. [PMID: 17053109]. variant of mitochondrial protein LOC387715/ARMS2, not HTRA1, is strongly associated with age-related macular 18. Dewan A, Liu M, Hartman S, Zhang SS, Liu DT, Zhao C, Tam degeneration. Proc Natl Acad Sci USA 2007; 104:16227-32. PO, Chan WM, Lam DS, Snyder M, Barnstable C, Pang CP, [PMID: 17884985]. Hoh J. HTRA1 promoter polymorphism in wet age-related macular degeneration. Science 2006; 314:989-92. [PMID: 29. Hadley D, Orlin A, Brown G, Brucker AJ, Ho AC, Regillo 17053108]. CD, Donoso LA, Tian L, Kaderli B, Stambolian D. Analysis

331 Molecular Vision 2017; 23:318-333 © 2017 Molecular Vision of six genetic risk factors highly associated with AMD in 39. Gibbs D, Yang Z, Constantine R, Ma X, Camp NJ, Yang X, the region surrounding ARMS2 and HTRA1 on chromosome Chen H, Jorgenson A, Hau V, Dewan A, Zeng J, Harmon 10, region q26. Invest Ophthalmol Vis Sci 2010; 51:2191-6. J, Buehler J, Brand JM, Hoh J, Cameron DJ, Dixit M, Tong [PMID: 19933195]. Z, Zhang K. Further mapping of 10q26 supports strong 30. Arakawa S, Takahashi A, Ashikawa K, Hosono N, Aoi T, association of HTRA1 polymorphisms with age-related Yasuda M, Oshima Y, Yoshida S, Enaida H, Tsuchihashi macular degeneration. Vision Res 2008; 48:685-9. [PMID: T, Mori K, Honda S, Negi A, Arakawa A, Kadonosono K, 18207215]. Kiyohara Y, Kamatani N, Nakamura Y, Ishibashi T, Kubo M. 40. Fritsche LG, Chen W, Schu M, Yaspan BL, Yu Y, Thorleifsson Genome-wide association study identifies two susceptibility G, Zack DJ, Arakawa S, Cipriani V, Ripke S, Igo RP Jr, loci for exudative age-related macular degeneration in the Buitendijk GH, Sim X, Weeks DE, Guymer RH, Merriam JE, Japanese population. Nat Genet 2011; 43:1001-4. [PMID: Francis PJ, Hannum G, Agarwal A, Armbrecht AM, Audo I, 21909106]. Aung T, Barile GR, Benchaboune M, Bird AC, Bishop PN, 31. Gotoh N, Yamashiro K, Nakanishi H, Saito M, Iida T, Branham KE, Brooks M, Brucker AJ, Cade WH, Cain MS, Yoshimura N. Haplotype analysis of the ARMS2/HTRA1 Campochiaro PA, Chan CC, Cheng CY, Chew EY, Chin KA, region in Japanese patients with typical neovascular age- Chowers I, Clayton DG, Cojocaru R, Conley YP, Cornes BK, related macular degeneration or polypoidal choroidal Daly MJ, Dhillon B, Edwards AO, Evangelou E, Fagerness vasculopathy. Jpn J Ophthalmol 2010; 54:609-14. [PMID: J, Ferreyra HA, Friedman JS, Geirsdottir A, George RJ, 21191724]. Gieger C, Gupta N, Hagstrom SA, Harding SP, Haritoglou C, Heckenlively JR, Holz FG, Hughes G, Ioannidis JP, Ishi- 32. Richardson AJ, Islam FM, Aung KZ, Guymer RH, Baird PN. bashi T, Joseph P, Jun G, Kamatani Y, Katsanis N. C NK, An intergenic region between the tagSNP rs3793917 and Khan JC, Kim IK, Kiyohara Y, Klein BE, Klein R, Kovach rs11200638 in the HTRA1 gene indicates association with JL, Kozak I, Lee CJ, Lee KE, Lichtner P, Lotery AJ, Meit- age-related macular degeneration. Invest Ophthalmol Vis inger T, Mitchell P, Mohand-Said S, Moore AT, Morgan DJ, Sci 2010; 51:4932-6. [PMID: 20445115]. Morrison MA, Myers CE, Naj AC, Nakamura Y, Okada Y, 33. Tian J, Yu W, Qin X, Fang K, Chen Q, Hou J, Li J, Chen D, Orlin A, Ortube MC, Othman MI, Pappas C, Park KH, Pauer Hu Y, Li X. Association of genetic polymorphisms and age- GJ, Peachey NS, Poch O, Priya RR, Reynolds R, Richardson related macular degeneration in Chinese population. Invest AJ, Ripp R, Rudolph G, Ryu E, Sahel JA, Schaumberg DA, Ophthalmol Vis Sci 2012; 53:4262-9. [PMID: 22618592]. Scholl HP, Schwartz SG, Scott WK, Shahid H, Sigurdsson H, Silvestri G, Sivakumaran TA, Smith RT, Sobrin L, Souied 34. Tam PO, Ng TK, Liu DT, Chan WM, Chiang SW, Chen LJ, EH, Stambolian DE, Stefansson H, Sturgill-Short GM, DeWan A, Hoh J, Lam DS, Pang CP. HTRA1 variants in Takahashi A, Tosakulwong N, Truitt BJ, Tsironi EE, Uitter- exudative age-related macular degeneration and interactions linden AG, van Duijn CM, Vijaya L, Vingerling JR, Vithana with smoking and CFH. Invest Ophthalmol Vis Sci 2008; EN, Webster AR, Wichmann HE, Winkler TW, Wong TY, 49:2357-65. [PMID: 18316707]. Wright AF, Zelenika D, Zhang M, Zhao L, Zhang K, Klein 35. Akagi-Kurashige Y, Yamashiro K, Gotoh N, Miyake M, ML, Hageman GS, Lathrop GM, Stefansson K, Allikmets Morooka S, Yoshikawa M, Nakata I, Kumagai K, Tsujikawa R, Baird PN, Gorin MB, Wang JJ, Klaver CC, Seddon JM, A, Yamada R, Matsuda F, Saito M, Iida T, Sugahara M, Pericak-Vance MA, Iyengar SK, Yates JR, Swaroop A, Weber Kurimoto Y, Cheng CY, Khor CC, Wong TY, Yoshimura BH, Kubo M, Deangelis MM, Leveillard T, Thorsteinsdottir N. Nagahama Cohort Research G. MMP20 and ARMS2/ U, Haines JL, Farrer LA, Heid IM, Abecasis GR, Consortium HTRA1 Are Associated with Neovascular Lesion Size in AMDG. Seven new loci associated with age-related macular Age-Related Macular Degeneration. Ophthalmology 2015; degeneration. Nat Genet 2013; 45:433-9. . 122:2295-302. . 41. Yamada Y, Tian J, Yang Y, Cutler RG, Wu T, Telljohann RS, 36. Rivera A, Fisher SA, Fritsche LG, Keilhauer CN, Lichtner Mattson MP, Handa JT. Oxidized low density lipoproteins P, Meitinger T, Weber BH. Hypothetical LOC387715 is a induce a pathologic response by retinal pigmented epithelial second major susceptibility gene for age-related macular cells. J Neurochem 2008; 105:1187-97. [PMID: 18182060]. degeneration, contributing independently of complement 42. Munoz-Erazo L, Natoli R, Provis JM, Madigan MC, King NJ. factor H to disease risk. Hum Mol Genet 2005; 14:3227-36. Microarray analysis of gene expression in West Nile virus- [PMID: 16174643]. infected human retinal pigment epithelium. Mol Vis 2012; 37. Yu W, Dong S, Zhao C, Wang H, Dai F, Yang J. Cumulative 18:730-43. [PMID: 22509103]. association between age-related macular degeneration and 43. Teper SJ, Nowinska A, Wylegala E. A69S and R38X ARMS2 less studied genetic variants in PLEKHA1/ARMS2/HTRA1: and Y402H CFH gene polymorphisms as risk factors for a meta and gene-cluster analysis. Mol Biol Rep 2013; neovascular age-related macular degeneration in Poland 40:5551-61. [PMID: 24013816]. - a brief report. Med Sci Monit 2012; 18:PR1-3. [PMID: 38. Deangelis MM, Ji F, Adams S, Morrison MA, Harring AJ, 22293892]. Sweeney MO, Capone A Jr, Miller JW, Dryja TP, Ott J, Kim 44. Minor EA, Court BL, Dubovy S, Wang G. AMD-associated IK. Alleles in the HtrA serine peptidase 1 gene alter the risk variants at the chromosome 10q26 locus and the stability of neovascular age-related macular degeneration. Ophthal- mology 2008; 115:1209-15. .

332 Molecular Vision 2017; 23:318-333 © 2017 Molecular Vision of ARMS2 transcripts. Invest Ophthalmol Vis Sci 2013; 52. Kanda A, Stambolian D, Chen W, Curcio CA, Abecasis GR, 54:5913-9. [PMID: 23942973]. Swaroop A. Age-related macular degeneration-associated 45. Friedrich U, Myers CA, Fritsche LG, Milenkovich A, Wolf A, variants at chromosome 10q26 do not significantly alter Corbo JC, Weber BH. Risk- and non-risk-associated variants ARMS2 and HTRA1 transcript levels in the human retina. at the 10q26 AMD locus influence ARMS2 mRNA expres- Mol Vis 2010; 16:1317-23. [PMID: 20664794]. sion but exclude pathogenic effects due to protein deficiency. 53. Chowers I, Meir T, Lederman M, Goldenberg-Cohen N, Cohen Hum Mol Genet 2011; 20:1387-99. [PMID: 21252205]. Y, Banin E, Averbukh E, Hemo I, Pollack A, Axer-Siegel R, 46. Micklisch S, Lin Y, Jacob S, Karlstetter M, Dannhausen Weinstein O, Hoh J, Zack DJ, Galbinur T. Sequence vari- K, Dasari P, von der Heide M, Dahse HM, Schmolz L, ants in HTRA1 and LOC387715/ARMS2 and phenotype and Grassmann F, Alene M, Fauser S, Neumann H, Lorkowski response to photodynamic therapy in neovascular age-related S, Pauly D, Weber BH, Joussen AM, Langmann T, Zipfel macular degeneration in populations from Israel. Mol Vis PF, Skerka C. Age-related macular degeneration associated 2008; 14:2263-71. [PMID: 19065273]. polymorphism rs10490924 in ARMS2 results in deficiency 54. Wang G, Dubovy SR, Kovach JL, Schwartz SG, Agarwal A, of a complement activator. J Neuroinflammation 2017; 14:4- Scott WK, Haines JL, Pericak-Vance MA. Variants at chro- [PMID: 28086806]. mosome 10q26 locus and the expression of HTRA1 in the 47. An E, Sen S, Park SK, Gordish-Dressman H, Hathout Y. Iden- retina. Exp Eye Res 2013; 112:102-5. [PMID: 23644223]. tification of novel substrates for the serine protease HTRA1 55. Wang G, Scott WK, Haines JL, Pericak-Vance MA. Genotype in the human RPE secretome. Invest Ophthalmol Vis Sci at polymorphism rs11200638 and HTRA1 expression level. 2010; 51:3379-86. [PMID: 20207970]. Arch Ophthalmol 2010; 128:1491-3. [PMID: 21060055]. 48. Iejima D, Nakayama M, Iwata T. HTRA1 Overexpression 56. Hara K, Shiga A, Fukutake T, Nozaki H, Miyashita A, Induces the Exudative Form of Age-related Macular Degen- Yokoseki A, Kawata H, Koyama A, Arima K, Takahashi eration. J Stem Cells 2015; 10:193-203. [PMID: 27125063]. T, Ikeda M, Shiota H, Tamura M, Shimoe Y, Hirayama 49. Iejima D, Itabashi T, Kawamura Y, Noda T, Yuasa S, Fukuda M, Arisato T, Yanagawa S, Tanaka A, Nakano I, Ikeda S, K, Oka C, Iwata T. HTRA1 (high temperature requirement Yoshida Y, Yamamoto T, Ikeuchi T, Kuwano R, Nishizawa A serine peptidase 1) gene is transcriptionally regulated by M, Tsuji S, Onodera O. Association of HTRA1 mutations insertion/deletion nucleotides located at the 3′ end of the and familial ischemic cerebral small-vessel disease. N Engl ARMS2 (age-related maculopathy susceptibility 2) gene in J Med 2009; 360:1729-39. [PMID: 19387015]. patients with age-related macular degeneration. J Biol Chem 57. Cheung CM, Wong TY. Is age-related macular degeneration a 2015; 290:2784-97. [PMID: 25519903]. manifestation of systemic disease? New prospects for early 50. Chan CC, Shen D, Zhou M, Ross RJ, Ding X, Zhang K, Green intervention and treatment. J Intern Med 2014; 276:140-53. WR, Tuo J. Human HtrA1 in the archived eyes with age- [PMID: 24581182]. related macular degeneration. Trans Am Ophthalmol Soc 58. Grassmann F, Heid IM, Weber BH. Recombinant Haplotypes 2007; 105:92-7. , discussion 7–8.. [PMID: 18427598]. Narrow the ARMS2/HTRA1 Association Signal for Age- 51. Tuo J, Ross RJ, Reed GF, Yan Q, Wang JJ, Bojanowski CM, Related Macular Degeneration. Genetics 2016; [PMID: Chew EY, Feng X, Olsen TW, Ferris FL 3rd, Mitchell P, 27879347]. Chan CC. The HtrA1 promoter polymorphism, smoking, 59. Oliver VF, Jaffe AE, Song J, Wang G, Zhang P, Branham KE, and age-related macular degeneration in multiple case- Swaroop A, Eberhart CG, Zack DJ, Qian J, Merbs SL. Differ- control samples. Ophthalmology 2008; 115:1891-8. [PMID: ential DNA methylation identified in the blood and retina 18718667]. of AMD patients. Epigenetics 2015; 10:698-707. [PMID: 26067391].

Articles are provided courtesy of Emory University and the Zhongshan Ophthalmic Center, Sun Yat-sen University, P.R. China. The print version of this article was created on 14 June 2017. This reflects all typographical corrections and errata to the article through that date. Details of any changes may be found in the online version of the article.

333